Introduction to Quantitative Causal Inference
The formal approach to statistical methods that address causal questions, known as ‘causal inference’, is increasingly seen as the way forward for quantitative epidemiology, with the concepts and methods contained fast becoming vital tools in any quantitative researcher’s box.
In this one-day course, we will introduce the key foundational concepts and tools in causal inference (potential outcomes, directed acyclic graphs) before moving on to see the benefits of these when applied to the study of mediation, i.e. questions concerning causal mechanisms such as “how much of the effect of socio-economic status on breast cancer survival is mediated by adherence to screening?”.
Although the formal nature of the subject inevitably requires that some mathematical details be given, the emphasis here will be on the main concepts and their application in a few realistic examples.
The important concepts in formal causal inference thinking will be introduced, along with some practical examples of this thinking guiding the analysis of data.
Some of the methods introduced will be used in examples using Stata or R (participants will be able to choose which of the two packages to use in the practical sessions).
|09:00-09:30||Registration and coffee|
|09:30-10:30||Introduction to causal concepts I: what is your causal question?||Rhian Daniel|
|10:45-11:30||Practical: Introduction to causal concepts II: Assumptions and Directed Acyclic Graphs||Rhian Daniel, Michail Katsoulis|
|11:30-12:45||Causal concepts||Rhian Daniel|
|14:00-14:45||Causal mediation analysis I: estimands and methods||Rhian Daniel|
|15:00-15:45||Causal mediation analysis II: examples||Rhian Daniel|
|15:45-17:00||Practical: Causal mediation analysis||Rhian Daniel, Michail Katsoulis|
- Dr Rhian Daniel (Lead Tutor)
Rhian studied mathematics as an undergraduate at Cambridge before spending twelve years at the London School of Hygiene & Tropical Medicine, first as an MSc student of medical statistics, then as a PhD student in missing data methods under the supervision of Mike Kenward, and most recently as an MRC Biostatistics and Wellcome Trust Sir Henry Dale fellow focusing on methods for statistical causal inference.
In March 2017, Rhian starts a new position at the Farr Institute London and UCL, where she will continue her Henry Dale fellowship. The aims of this fellowship are to extend causal mediation analysis to handle the high dimensional mediators encountered in modern biomedical contexts, including those arising from OMICs technologies.
Rhian collaborates with Juan Pablo Casas on OMICs data, and with Bianca De Stavola and Stijn Vansteelandt on mediation and causal inference methods.
- Dr Michail Katsoulis
Michail holds a BSc in Applied Mathematical and Physical Sciences (2006), an MSc in Statistics (2007) and a PhD in Epidemiology (2015).
He worked at the Hellenic Health Foundation and the University of Athens for 7 years (2008-2015) before joining the Farr Institute of Health Informatics Research London as a Medical Statistician (Research Associate) in December 2015.
His research interests include interaction and mediation analysis, Mendelian randomisation and application of causal inference methods to deal with time-dependent confounding.